Summary The Boeing 737-210C (C-GNWN, serial number21067) was operating as Flight FAB6501 from Lupin, Nunavut, to Edmonton, Alberta. The runway visual range (RVR) provided to the flight crew prior to commencing the approach to Runway12 at Edmonton was 1200RVR, with a runway light setting of5. The crew flew the instrument landing system approach in darkness and touched down on the infield to the left of the runway surface, at 0544 mountain standard time. The aircraft travelled approximately 1600feet before returning to the runway. After the aircraft was brought to a full stop, aircraft rescue and firefighting was requested by the flight crew. One runway light, four taxiway lights, and one hold sign were struck by the aircraft. There were no injuries and the passengers deplaned via the rear airstairs door. Ce rapport est galement disponible en anglais. Other Factual Information Records show that the aircraft was certified and equipped, and was maintained in accordance with existing regulations and approved procedures. C-GNWN is a Boeing737-210C (combi) that was modified with a forward main cabin cargo compartment, a cargo door and a gravel kit. The aircraft's weight and centre of gravity were within limits. Testing of the aircraft's auto-flight and approach avionics revealed the equipment was operating within design tolerances. The aerodrome forecast for the airport, issued on 25February2004 at 0439 mountain standard time (MST)1 and valid from 0500 to 0500, was as follows: winds variable at 3knots, visibility 0statute miles (sm) in freezing fog, temporarily from 0500 to 0800, 2sm in mist, and ceilings broken at 25000feet above sea level. The hourly weather report (METAR) issued at 0600, 15minutes after the occurrence, was as follows: winds calm, visibility 1/8sm, Runway 12RVR 1000feet, Runway30RVR 800feet in freezing fog, vertical visibility 300feet, temperature -14 C, dew point -15C, altimeter setting29.76, and remarks, fog8/8and frost on the indicator. The flight crew members were certified and qualified for the flight in accordance with existing regulations. The captain had a total of 25000hours total time with about 6000hours on B-737 aircraft. In the previous 30days, the captain flew 21hours. The first officer had approximately 6000hours total time with about 3200hours on B-737 aircraft. In the previous 30days, the first officer flew 33 hours. Both pilots woke at approximately 0530 on February24, to start the schedule. They had obtained adequate rest before the start of the schedule and were not tired when they checked into the day-room in Yellowknife. Neither pilot slept for any significant amount of time in the day-room, having just wakened a few hours earlier. The First Air Operations Manual, Section4.1.3.3, which is compliant with Canadian Aviation Regulations, allows flight crews to reset their duty day if they are provided an opportunity to obtain not less than eight consecutive hours of sleep in suitable accommodation, time for meals, personal hygiene, and time to travel to and from the rest facility. The flight crew arrived at the hotel in Yellowknife at about 1000 and left at about1830. The captain and first officer had been awake for almost all of the 24hours before the occurrence. Studies show that long periods of wakefulness produce a significant degradation in human performance.23 Performance decrements are noted in reaction time, arithmetic, signal detection, meter reading accuracy, safety alarm alerts, random number addition speed, psychomotor and cognitive functioning, etc.4 Rest taken during a person's normal waking period is not normally of sufficient quality to be fully restorative.5 These rests may reduce fatigue, but not all of the performance decrements associated with residual fatigue.6 However, even a full eight hours of sleep would not be sufficient to re-set the crew's circadian rhythm for a flight planned late at night.7 FAB 6501 was the return flight from Lupin (CYWO) on a charter that started in Edmonton (CYEG) the previous evening. This was a relatively new charter contract for a mining company, and this was the third time First Air had conducted the trip. The schedule for the charter required the flight crew to fly four legs over a 20-hour period with a 10-hour rest period during the daytime stopover in Yellowknife, Northwest Territories (CYZF). The scheduled arrival and departure times for the flights were planned as follows, starting on 24February2004: FAB 951 CYEG - CYZF depart 0725 arrive 0912 (day room provided inYZF) FAB 956 CYZF - CYEG depart 2045 arrive 2226 FAB 6500 CYEG - CYWO depart 2345 arrive 0145 FAB 6501 CYWO - CYEG depart 0305 arrive 0505 The aircraft's actual arrival and departure times were fairly close to the scheduled times until the departure from Lupin. The flight crew was advised by First Air dispatch of low visibilities due to fog in Edmonton, and they delayed the departure for approximately 40minutes to assess the weather and plan for contingencies. FAB 6501 departed Lupin at 0340 MST with Calgary, Alberta, as the alternate airport. When FAB 6501 contacted Edmonton arrival at 0519, the visibility for Runway30 was 1200RVR, the approach ban limit, and the flight crew briefed for an approach to that runway. During the next 14minutes, the visibility for Runway30 increased to 3000RVR but then decreased to 900RVR with the other runways, 12and02, reporting 800RVR. The flight crew planned to hold for approximately 30minutes at an initial approach fix for Runway30 and wait for the visibility to increase. Before entering the hold, the visibility for Runway12 increased to 1200RVR with a light setting of5, so the crew requested the instrument landing system (ILS) approach to Runway12 and briefed for that approach. During radar vectors for the approach, the visibility for Runway 12 remained at 1200 RVR, while Runway30 averaged 600RVR. At 0541, FAB6501 switched to tower frequency and was given the following information: wind 220Magnetic(M) at 3knots, altimeter setting29.76, and 1200RVR with a light setting of5. Less than one minute later, FAB6501 crossed the final approach fix (Devon beacon) established on the ILS glide path and localizer for Runway12. The first officer was flying the approach and briefed for a pilot-monitored approach (PMA). The generic industry PMA technique used by some companies is one where the first officer flies the aircraft on to the approach minimums while the captain monitors the instruments. Approaching minimums, the captain begins to look outside for the appropriate visual cues. At decision height, if the captain has the appropriate visual cues to land, he would take control, or have the first officer continue on instruments until the appearance of more visual cues and then take control, and complete the visual landing. When control is transferred, the first officer continues to monitor the flight instruments until touchdown. If visual references are lost at any time, the captain would command a missed approach, and the first officer would fly the missed approach procedure. Both flight crew members of FAB 6501 had worked for another employer flying Boeing737s with a PMA procedure similar to the generic industry PMA standard operating procedure (SOP). The flight crew used the PMA SOPs from their previous employer rather than the First Air PMA SOPs, as they felt they were more complete and were better suited to the conditions in which they were flying. Autopilot Approach Modes VOR/LOC mode is used to automatically intercept and track the selected radio course. Crosswind compensation occurs after the course is engaged. AUTO APP mode is used to automatically capture ILS localizer and glide slope. After glide path capture at 1500 feet RAD ALT or below, localizer sensitivity is reduced from 100% to 50% as altitude decreases to 100 feet. Glide path sensitivity is reduced to 0% as altitude decreases to 50 feet. This attenuation feature is incorporated to prevent the aircraft from making any large pitch or roll changes low to the ground. There is no localizer sensitivity attenuation while in VOR/LOC mode. The First Air PMA SOP says that the captain shall monitor the flight instruments and ensure that the procedure is being flown as briefed. The first officer shall fly the aircraft and published approach as briefed, until the captain assumes control. At decision height, if the captain has sufficient visual references to land the aircraft, he shall announce, minimums - my controls, landing. The first officer will respond your controls and assume pilot-not-flying duties. When the captain has assumed control, the first officer shall continue to monitor the flight instruments until the aircraft has landed and reverse thrust has been initiated. Autopilot Approach Modes VOR/LOC mode is used to automatically intercept and track the selected radio course. Crosswind compensation occurs after the course is engaged. AUTO APP mode is used to automatically capture ILS localizer and glide slope. After glide path capture at 1500feet RAD ALT or below, localizer sensitivity is reduced from 100% to 50% as altitude decreases to 100feet. Glide path sensitivity is reduced to 0% as altitude decreases to 50feet. This attenuation feature is incorporated to prevent the aircraft from making any large pitch or roll changes low to the ground. There is no localizer sensitivity attenuation while in VOR/LOC mode. When FAB6501 intercepted the final approach course, the autopilot was set in VOR/LOC mode. The First Air Boeing737 Operations Manual stated that if glide slope signals were erratic, only LOC mode was to be used on the autopilot. This was the preferred method used by First Air flight crews to avoid excessive pitch movements by the autopilot when coupled to the glide path in AUTO APP mode. These movements could be the result of obstacles interfering with the glide path, autopilot anomalies, or configuration changes requiring a lot of trim adjustment - lowering flaps and landing gear. During the approach, just prior to crossing the final approach fix, the aircraft configuration was changed, in about 30seconds, from flaps5 and landing gear up to flaps30 and gear down. While in VOR/LOC mode, the glide path was being followed in control wheel steering mode, where the first officer manually controlled the pitch of the aircraft through the autopilot to maintain the glide path profile. After crossing the final approach fix, with aircraft configured for landing, the first officer elected to leave the autopilot in VOR/LOC mode. The aircraft was in clear air until approximately 500feet above ground level (agl), when it entered the ground-based fog layer. The glow of the runway approach lighting was observed by both the load master, seated in the jump seat, and the captain. At minimums (200feetagl) the captain saw the approach lighting strobes and commanded the first officer to continue. At approximately 65feetagl, the captain observed the runway edge lighting, took control of the aircraft, disconnected the autopilot and reduced the rate of descent. The first officer continued to monitor the instruments and called out airspeeds and descent rates. The aircraft was equipped with a flight data recorder (FDR) and cockpit voice recorder (CVR). Both operated as expected and provided useable information. The FDR data showed that the aircraft crossed the threshold at a height of approximately 45feet at 129KIAS (Vref+10), heading 120M. Descending over the runway, manual control inputs resulted in the aircraft drifting slightly right of the runway centreline, with a maximum displacement of approximately 10feet. At about 40feet agl, the aircraft was turned to the left to regain the centreline; the maximum bank angle achieved was 16left wing down during this manouevre before returning to a near wings level attitude. This would have placed the left wing tip 33feet and left engine pod 31feet above the runway surface. At 15feet agl, the power was reduced to idle; the captain had difficulty seeing the runway edge lighting at that time. Approximately four seconds later, it became evident to the flight crew that the aircraft was going to touch down beside the runway. The first officer called for go-around thrust and started to advance the thrust levers. Recognizing the aircraft's low energy state and position relative to the runway, the captain immediately retarded the thrust levers, and the aircraft touched down with a vertical decelerationof2.3g. At touchdown the aircraft's heading was 117M. The right main landing gear touched down first, approximately 8feet off the left edge of the runway and 2700feet from the threshold. The initial track of the main gear through the snow was approximately 109M. Reverse thrust and a large input of right rudder were applied shortly after touchdown. The marks left by the aircraft through the snow were consistent with the aircraft sliding sideways, up to 23nose-right of the aircraft's track. These corrective actions by the flight crew resulted in the aircraft regaining the runway after travelling 1600feet beside it and crossing TaxiwayA2. The aircraft was brought to a stop on the runway centreline, 4550feet from the threshold (seeFigure1). At the time of touchdown, the RVR was 1200for Runway12 and 800forRunway30. Figure1. Overhead view of Runway 12 at CYEG During the aircraft's rollout beside the runway, the nose gear gravel deflector detached, damaging the electrical/avionics bay doors and structure directly behind it. In addition, there were numerous areas of puncture and denting damage to the underside fuselage skin surface and a large puncture on the keel beam area of the belly just aft of the main gear. The left engine contacted and partially ingested a HOLD sign located on the west side of TaxiwayA2 and both engines ingested grass and dirt. Aircraft rescue and firefighting (ARFF) personnel responded to the crash alarm from the tower and began making their way out to the aircraft at 0545. Visibility was described as anywhere from 100to 300feet, and their progress was impeded by the fog. The first ARFF vehicle arrived at the aircraft with the assistance of forward-looking infrared camera equipment, four minutes after the alarm was activated. Runway12 at CYEG is 10200feet long and equipped with the required lighting and markings for a Category1 ILS runway, in accordance with Transport Canada TP312E. The lighting system comprised white parallel runway edge lights, green threshold lights, red runway end lights, 1000 feet of sequenced flashing approach lights, and 1400feet of white runway alignment indicator lights. This system was described as a simplified short approach lighting system with runway alignment lights (SSALR). It has five intensity settings. AppendixA shows the differences between Categoryl and Categoryll approaches and runway lighting systems. The only difference in the approach phase is the lower decision height for a Categoryll approach. The main difference between the two is the approach and runway lighting requirements. It is clear from AppendixA that the lighting required for a Categoryll approach is far superior to that required for a Categoryl approach. In particular, the approach lighting is massive and two-dimensional, allowing a pilot to more easily keep the aircraft wings level and clearly determine the centre of the approach path. As well, the first 3000feet of the runway has a two-dimensional lighting system, with centreline and touchdown zone lighting in addition to the runway edge lights. Canadian regulations permit Categoryl approaches to be conducted in weather conditions equivalent to or lower than Categoryll landing minima. However, the air and ground equipment requirements, procedural requirements and crew requirements for conducting Categoryll approaches are much more stringent.8 To the flight crew, after the aircraft stopped on the runway, the runway edge lighting did not appear bright enough for light setting5. The light setting for a given runway is not recorded; however, information gathered during the investigation supports the conclusion that most probably the lights were on setting 5for the approach and landing, and for a time after the occurrence to facilitate ARFF and airport vehicle movements to and from the aircraft. Maintaining airport lighting brightness or conspicuousness is achieved by ensuring that the power generating equipment is producing the necessary power, that the circuit that delivers the power is in good condition, and that the lights are operable and clear of contamination. The runway lighting system is maintained by the Edmonton Regional Airport Authority (ERAA). Inspection records showed that the visual aid facilities were being monitored and maintained on a scheduled basis. These facilities included the actual lights and support structures, as well as lighting circuits and power generation equipment. A Transport Canada inspection of the approach and runway edge lighting for Runway12 was completed on 15January2004, and the lighting conformed to the applicable standards in TP312; however, there were no Transport Canada standards with which to compare the lighting circuit and power generation measurements. The investigation used Federal Aviation Administration Advisory Circular150/5340-26 to obtain examples of suggested maintenance practices and equipment tolerances for comparison. An independent audit organized by the ERAA determined that the visual aid facilities were being maintained to industry standards. A TSB Engineering Branch test, performed on 15June2004, showed that the runway edge lights for Runway12 were producing the required lighting intensity for light setting5. Two runway edge lights tested were not aligned to within of the runway direction as stipulated in TP312. The effect on overall luminosity caused by this misalignment is unknown. Post-occurrence inspection of the glide path and localizer antennas and equipment rooms indicated that all systems were operational with no faults found or indicated at any time during the approach or after touchdown. TSB investigators viewed the antennas and equipment rooms approximately two hours after the occurrence and no anomalies were noted. On 27February2004, NAV CANADA conducted a flight inspection for the ILS Runway12 at CYEG. The report concluded that the ILS (both localizer and glide path) was within tolerances as described by International Civil Aviation Organization standards. The maximum localizer deviation recorded at minimums (200feet agl, 0.55nautical miles from the threshold) was 7A (microamps), which equated to 21.3feet to the left of centreline.